Recent increases in opioid analgesic prescribing in the US and worldwide have coincided with increases in opioid-related morbidity and mortality. As described in a 2015 NIH workshop report and in multiple commentaries in peer-reviewed health journals, however, the exact harms of prescription opioid analgesics- and their precise role in the ongoing opioid epidemic-remain unclear. This uncertainty has been driven by the limited sample sizes and follow-up durations of randomized, controlled trials, as well as the challenge of ruling out biased selection and other confounding in observational studies. Notably, patients who receive prescription opioids may also be more likely to have pre-existing conditions such as substance use disorders. This process is referred to as adverse selection, and it suggests that prior estimates of opioid harms from some observational studies may have been artificially inflated. The overall objective of the proposed research is, therefore, to use advance pharmacoepidemiologic approaches to determine the extent to which prescribed opioids have behavioral harms. In particular, the proposed studies will estimate the risks of prescribed opioids for substance use disorders, motor vehicle accidents, depression, and suicidal behavior, each of which has been identified in prior literature as a potential opioid harm. In 2 population-level datasets-including 1.9 million opioid recipients drawn from the whole Swedish population and 26 million opioid recipients drawn from US insurance claims-these studies will evaluate the central hypothesis that prescribed opioids are associated with behavioral harms, which would be consistent with true adverse effects. First, the proposed research will examine the extent to which patients receive prescription opioids more frequently in the presence of pre- existing substance use disorders, mood disorders, and other social and behavioral factors. Having identified these potential confounds, analyses will subsequently estimate risk associations between opioid prescriptions and both short-term and long-term behavioral harms. Specifically, analyses will use within-person comparisons, which implicitly rule out all sources of time-invariant confounding, and time-varying statistical covariates, which explicitly rule out sources o time-varying confounding. The rationale for this approach is that the combination of within-person analyses, time-varying covariates, and data from two populations will help generate more accurate estimates of the adverse effects of prescribed opioids. The proposed research will be complemented with training in (a) pharmacoepidemiology and other quasi-experimental research, (b) clinical pain research, (c) collaborative science, and (d) the responsible conduct of research. Upon the completion of the proposed activities, the applicant is expected to have reached the long-term goal of establishing an independent program researching the causes and consequences of opioid and other substance use.